Dawei Gao and Xiaorui Sun tackles fairly the issue of harmonics elimination, unbalanced problems and power factor compensation. This work presents a three phase shunt active power filter with control method based on the traditional neural network algorithm, extracting the fundamental sinusoid from distorted load current waveform without harmfully phase shift due to adapting electronic bandpass filters. The control circuit is used in generating the reference current of the compensation current also controls the PWM voltage source inverter to produce the compensation current following the reference current. These authors employed a number of computer simulations with different operating conditions using MATLAB/SIMULINK environment in verifying the performance of the control method and the shunt active power filter. In conclusion, these methods have been analyzed and the active power filter can eliminate harmonics, compensate power factor, and correct unbalanced problems simultaneously. Also, the THD figure is found within the 5% range of the IEEE - 519 standard but not as perfect as that for the modified adaptive artificial neural network discussed above.
In [p2], a novel control design of shunt active power filter to compensate reactive power and reduce the unwanted harmonics is adopted. A shunt active filter is realized employing three phase voltage source inverter (VSI) and a control circuit. A modified adaptive neural network (ANN) controller is design to estimate the harmonic components of the distorted load current and supply voltage. A power factor correction function is incorporated in the shunt active filter to achieve a power factor that is near to unity. Different cases are considered by the author and then simulated in order to show the validity of the active power filter with the conventional neural network control. The voltage source inverter consists of IGBT switches and an energy storage capacitor on Dc bus is implemented as a shunt active power filter. Both this are employed using the W-H algorithm. The presented simulation result obtained at the end of Wenjin Dai Yu Wang work is realized using MATLAB/SIMULINK power system toolbox for a 3-É¸, 4-wire power distributed system with a shunt active power filter. The compensation current has compensated the random varying current with a drastic increase in the firing angle α from 25° to 45°, hence, source current THD% details obtained as below:
Without APF: At α= 25°, THD%= 24.10% and at α= 45°, THD%= 27.06%.
With APF: At α= 25°, THD%= 2.93% and at α= 45°, THD%=2.12%.
The above results shows how feasible and effective this thesis work is with all the THD values lying on an acceptable IEEE - 519 requirements.
Z. F. Hussien et al [p3] in their work presents the design and simulation of a 1-É¸ shunt active power filter for harmonic and power factor compensation of multiple non-linear loads. The system consist of an uncontrolled rectifier and ac controller as the non-linear loads, with an active filter to compensate for the harmonic current drawn by the load. The active filter is based on a full-bridge 1-É¸ inverter and its design controller is based on time-domain method that consists of three main tasks: to identify the harmonic current and form a synchronized reference, to provide closed-loop control to force the current of the active filter to flow the reference and to regulate the capacitor dc voltage. The authors also discussed the effects of varying the switching frequency on the performance of the active filter. The complete system has been modeled and simulated successfully in MATLAB/SIMULINK with the following simulation results. It was observed that the performance of the active power filter is investigated as its switching frequency is varied. Performance for switching frequency from 5kHz to 25kHz, its THD decreases drastically as such, they conclude that the higher the switching frequency, the better the active power filter is to synthesize the reference current and make the supply current sinusoidal, but only this research presents a huge disadvantage of higher switching losses.
In [p4], A. Senthilkumar and Dr .P. Ajay-D-Vimal Raj explores the application of artificial intelligence on solving the power quality problems by using the shunt active power filter technique to 3-É¸, 3-wire distributed systems. A unit vector template generation control strategy is modeled as current controller for the shunt active power filter strategy. The proportional and integral (PI) controller is designed to minimize error between the actual and the reference DC voltage of the shunt active power filter strategy. The transient period and peak overshoot of DC bus voltage using a PI controller is observed to be higher in initial and load change conditions. The authors present a new neural learning algorithm (NAL) for the current controller of the shunt active power filter strategy. The performance of the presented neural learning algorithm (NAL) is extensively analyzed of diode rectifier RL non linear load with respect to two different operating conditions. The work presented by the authors is designed with MATLAB/SIMULINK environment and the following simulation results for before and after connecting shunt active power filter is given below.
Analysis of THD of a source voltage before and after connecting SAPF:
The dynamic response of the source voltage waveform for the non-linear load for before and after connecting the SAPF strategy is: Before, THD%= 6.76%, After, THD%= 0.79%, THD minimization, by using SAPF, 91.34% of source voltage THD is minimized.
Analysis of THD of a source current before and after connecting SAPF:
Before, THD%= 36.02%, After, THD%= 2.50%, THD minimization, by using SAPF, 94.07% of THD of source current is minimized.
In conclusion, we see how effective and reliable the new learning algorithm (NAL) is when employed for the current controller of the SAPF strategy for the minimization inherent power quality problems, with IEEE - 519 requirements obeyed. This test result also brings out the advantages of SAPF strategy for power quality problems management.
Nils Hoffmann et al [p5] in their work present a SAPF that uses an optimization algorithm for the harmonic reference current calculation. The algorithm is designed for an active filter that is initially rated for remediating only the harmonic currents dispense by the non-linear industrial loads. To realize this objective, the algorithm takes into consideration the following power quality indices: total harmonic distortion, individual harmonic distortion and power factor. This optimization algorithm is verified by simulating several cases of practical interest. The optimization algorithm is developed to reach important power quality indices at the point of common coupling (PCC) of an industrial plant. The optimal reference current calculation method makes it possible to compensate reactive and harmonic currents for an APF hardware initially rated for migration of only harmonic currents. The method uses Lagrange-Multiplier technique to derive the optimal achievable line-side current relative to the actual measured load current. The simulation is carried out and the result achieved power quality indices at the PCC are conforming to IEEE - 519 harmonics recommendations in all cases. The optimal reference current calculation offers limitations as: the optimization problem is quite complex in its formulation. It sometimes needs higher computational power to solve the resultant derivations online. It is associated with asymmetric outputs situations when calculations are implemented individually.
Reference [p6] discussed extensively the control of SAPF from two different aspects: synchronous detection method (SDM) and digital control based on instantaneous power theory (p-q theory). The authors in their work demonstrate the application of these methods to the control of APF. In the synchronous detection method (SDM), the authors assume that the 3-É¸ main currents are balanced after compensation, and it tries to determine that required amplitude of the mains currents. In the digital control based, the compensation technique of SAPF with the instantaneous reactive power in 3-É¸ circuits, namely the p-q theory allows dynamic power factor correction and both harmonics and zero-sequence current compensation. It describes a method of calculating the instantaneous power components from some transformed values of 3-É¸ voltages and currents. The algebraic transformation, known as Clark Transformation, converts the 3-É¸ a-b-c coordinates to α-β-φ coordinates. MATLAB/SIMULINK toolbox has used to develop the models to execute both the SDM and p-q theory calculations. In conclusion, this work shows that digital control provides better power quality improvement than the SDM.
In the work of M. A. Farahat, A. Zobah [p7], a new control design using an artificial neural network is used to make conventional shunt active adaptive filter. The control design is based on artificial neural networks that use a modified Widrow-Hoff weight-updating algorithm. The functionality of the shunt active filter is enhanced with the use of this artificial neural network algorithm. The shunt active power filter can compensate for balanced and unbalanced non-linear load currents, adapts itself to compensate for variations in non-linear load currents or non-linear load types and correct power factor of the supply side near unity. The general topology of the four-wire, three-leg active power filter is studied by the authors. The adaptive neural network system, self-charging circuit, hysteresis control and injection circuit are integrated as an overall model of adaptive shunt active filter. A complete model of the filtering system has been built by using MATLAB/SIMULINK simulation software. A computer simulation is carried out to verify the operating performance of the active filter. A 3-É¸ system is built using MATLAB. The simulation results shows that, the non-linear load current has a THD of 93.5%, but when the proposed adaptive shunt active filter is used, the current source has an overall THD of around 2.9%. In conclusion, active filters are an up-to-date solution to power quality problems. Simulation results and various system operating conditions have verified the operating performance of the proposed adaptive shunt active filter to be highly effective and robust. The proposed active filter has shown a noticeable good dynamics which fulfils its potential for real-time applications. The proposed active filter is effective to reduce the THD within the limit of standard IEEE - 519.
The work conducted by M. Rukonuzzaman and M. Nakaoka [p8] gives us another edge on 1-É¸ shunt active power filter with artificial neural network method for determining compensating current. An advanced active power filter is used for the compensation of instantaneous harmonic current components in nonlinear current loads. The authors incorporated the signal processing technique using an adaptive neural network algorithm for the detection of harmonic current components generated by nonlinear current loads and it can effectively determine the instantaneous harmonic components in real-time. The load current is sampled uniformly and one sample is taken at a time during the adaptive estimation of the harmonic current component by neural networks. These sampled values are then used in determining the magnitude and phase of the fundamental and harmonic current components via an adaptive neural circuit. The main aim of the algorithm is to keep the error at its global minimal. A hysteretic current controller is then employed to control the current in a voltage source inverter in such a way as to produce an output current that follows a reference current signal. A computer simulation is then carried out so as to verify the performance of the active power filter. The simulation results shows that, the THD is 23.34% before the harmonic compensation in the load current and 3.24% in the supply current after the harmonic current compensation in 1-É¸ diode bridge rectifier nonlinear current load as it conforms with the standard set aside by the IEEE - 519.
Another work conducted by Parag Kanjiya et al [p12] tekes a new control approach for the load compensation using 1-É¸ shunt active power filter (SAPF) under distorted supply condition in achieving maximum possible power factor, keeping total demand distortion (TDD) within the acceptable limit and making the source current waveform non-distinct as the supply voltage including all the harmonics of the voltage for unity power factor (UPF) operation. The authors describes a new control algorithm which ensures the best compromise between harmonic free (HF) and unity power factor (UPF) by keeping the total demand distortion (TDD) of the source current within the standard range. These two (HF and UPF) are described in these reference as the two main control strategies for load compensation under distorted supply condition. The HF strategy gives sinusoidal source currents, while for a demanded average load power UPF strategy gives minimum value of RMS source currents. Both UPF and HF control strategies will under ac supply results to same compensation performance indices. It should be clear that, HF and UPF operation cannot be realized simultaneously when the supply voltage is distributed. Also, UPF operation can be achieved only if the harmonics and profile of source current is exactly identical that of supply voltage. Hence, the control of SAPF using UPF control strategy will lead to a non-sinusoidal source current and as such violet the TDD limit set by the IEEE - 519 standards. So in order to realize this new approach in ensuring best compromise between HF and UPF, two PI (Proportional-Integral) controllers are used. One over DC link voltage is also employed which estimates the fundamental conductance factor with an over source current TDD which estimates the conductance factor corresponding to harmonics. A hysteretic current controller is incorporated for summing up the products of the conductance factor with fundamental and harmonic component of the supply voltage respectively in other to generate reference source current for the direct control of SAPF. The fundamental component is then extracted using Second Order Generalized Integration (SOGI) while the harmonics component is simply computed by subtracting the extracted fundamental component for measured voltage. MATLAB/SIMULINK environment using SIMPOWERSYSTEM block set has been used in verifying the performance of the control algorithm. The simulation result/performance achieved using conventional control algorithm (CCA) and proposed control algorithm (PCA) approach is given below:
So in conclusion, these new control algorithms which controls the SAPF under distorted supply condition gives us a good compromise performance between better power factor and sinusoidal current while keeping TDD of the source current within an acceptable limit. The compensation performance of this two (CCA and PAC) algorithms is compared. The authors found that, the increased in source power factor is magnified during lower loading condition and that by allowing the limited amount of supply voltage harmonics into the source current keeping TDD below acceptable limit, the source PF can be improved. It is finally clear that the proposed algorithm employed in this reference achieves best possible PF with limited source current TDD under distorted supply condition.
Recent publication in the area of work related to this thesis is reviewed in [p13]. The authors N F A Abdul Rahman and S. Z. Mohammad Noor in their work consider a new approach of single-phase shunt active power filter (SSAPF) operation. In contra with the conventional filters, the proposed filter acts as a new current pathway for the expected distorted current. From the Kirchhoff's Current Law (KCL), the result of the summation of the SSAPF consists of two sets of insulated gate base transistor (IGBT) and diode pair and the switching operation is controlled using unipolar pulse width modulator (UPWM) signal. The SSAPF with new unique operation is employed, unlike the conventional shunt active power filters (SAPFs), the filter acts as an AC-AC converter which forms an additional closed loop for current pathway so that when the electrical power system supplying a nonlinear load is live, the remaining current which complementary to the load current will flow in the SSAPF. Eventually, the summation of current from the SSAPF current and the load current results to a near fully sinusoidal shape and in phase with supply voltage. The complementary current flows via the two sets of (IGBT) and the diode pair. In this study, the unipolar pulse width modulation (UPWM) technique is applied to the IGBT. The control strategies for the new SAPF are presented which involve the UPWM signal generation. The signal is then used for controlling switching activities for both IGBTs. Several components such as (a peak detector, a sine wave generator, an absolute block, a product block, a subtractor, a proportional integral (PI) controller, a triangular or carrier generator and a comparator) are all employed engaging in producing the UPWM signals. Each component/block dispenses an important purpose in ensuring the UPWM signal is correctly and accurately generated. This work was simulated using MATLAB/SIMULINK (MLS) to verify the performance of the SSAPF. The simulation results focused on several aspects such as the operation of the proposed SSAPF, the connection point between the filter and the main electrical power system and the filters inductor value. The author's uses filters inductor values range between 0.4mH to 0.8mH and all parameters show less than 5% of THD. It was showed that, the higher the value of the inductor, the lower the THD%. Also, it was cleared to us that, the proposed SSAPF which connected before the line inductor show better result than the proposed SSAPF which connected after the line inductor. In addition to the test results, the proposed SSAPF which connected before the line inductor generates lower fundamental supply current than the proposed SSAPF which connected after the line inductor. So in essence, the proposed SSAPF has the full capability to improve the THD% regardless of its connection to the main electrical system. In conclusion, based on the simulation results, the proposed SSAPF which connected before the line inductor is much better as compared to the other connection.
In [p14], another development has been made, presenting a single-phase shunt active power filter (SSAPF) for current harmonic compensation based on neural filtering. A current controller inverter is employed in this study as the shunt active filter (SAF) which made it possible to compensate a nonlinear current load by receiving its reference from a neural adaptive notch filter. This is a recursive notch filter for the fundamental grid frequency (50Hz) and is based on the use of a linear adaptive neuron (ADALINE). This filters parameters are made adaptive with respect to the grid frequency fluctuation. Estimation of actual grid frequency and extracting the fundamental component from the coupling point voltage is realized by the use of a phase-locked loop system. A moltiresonant controller is also employed in the current control of the inverter. This estimated grid frequency is then fed to the neural adaptive filter and to the multiresonant controller and as such, the inverter creates a current equal in amplitude but opposite in sign to the load harmonic current, thus producing an almost sinusoidal grid current. The largest three harmonics of the load current to be compensated by the active power filter (APF) is realized by the use of an automatic turning of the multiresonant controller. All simulations have been done in a MATLAB/SIMULINK environment. The neural network (NN) based APF control has been compared both in numerical simulations and experimentally, with the classic APF control based on the P-Q- theory. These comparison correctly works and permitting a full compensation of the grid current harmonics, with no appreciable difference in the performance as shown below in the experimental test. In the frequency domain, the comparison between the results with both the control techniques does not present appreciable differences.
Compliance with the IEEE and IEC standard of the NN-based control (experiment):
Load 1: THD%= 3.62%, Load 2: THD%= 4.85%.
Compliance with the IEEE and IEC standard of the P-Q-theory based control (exp):
Load 1: THD%= 3.35%, Load 2: THD%= 4.75%.
Since the THD% in both cases is lower than the 5%, then the control techniques permits a full compliance with both IEC and IEEE standard.
The contribution of this work concluded by the authors that, besides the current harmonic compensation, the P-Q- theory approach permits one to make the system work at unity power factor, which the NN-based control does not. Similarly, with reference to the current controller, the multiresonant controller seems to be a good choice. It presents the significant advantage of separately controlling the fundamental current component to zero and some of its harmonics which the classic-hysteretic controller could not easily achieved expect by the implementation of higher switching frequency. Nevertheless, the multiresonant controller also possesses a drawback of limited harmonic control to the selected harmonics, in the chosen number or order.